Loading ToolBench
An MCP server that provides tools and resources for interacting with Google BigQuery, enabling SQL query execution and table listing.
Single tool with Japanese name 'BigQueryクエリ実行' lacks English verb-noun convention. Schema is present with proper Zod types (query, projectId, location, maxResults, params, dryRun), but description is minimal (11 words, 60 chars) and lacks context on when to use it, what it returns, or prerequisites. Parameters lack individual descriptions in the schema definition, only type info visible. No output schema documented. Error handling returns raw error messages without recovery guidance. Tool does one thing (execute query) but lacks composition guidance for multi-step workflows.
BigQueryに対してSQLクエリを実行します。
Tool name is in Japanese ('BigQueryクエリ実行') and lacks English verb-noun convention. LLMs expect English action verbs (execute_query, run_query). Non-English names reduce discoverability and LLM reasoning clarity.
Tool description is only 60 characters ('BigQueryに対してSQLクエリを実行します。'). Rubric requires 10 - 1024 chars with WHAT, WHEN, and prerequisites. Current description lacks context on return format, when to use vs alternatives, or error cases.
Input parameters (projectId, location, maxResults, params, dryRun) have no individual descriptions in the schema. Zod schema shows types but LLMs cannot infer meaning from names alone. 'params' is ambiguous, is it query parameters, request parameters, or substitution variables?
No output schema documented. Response returns JSON with 'rows' and 'metadata' fields, but LLMs cannot plan downstream calls without knowing field types, structure, or what metadata contains (jobId, totalBytesProcessed, cacheHit are present but undocumented).
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 36 | 2026-07-28+ | v2 |
Error handling returns raw error messages without recovery guidance. 'error instanceof Error ? error.message : String(error)' provides no actionable next steps. Rubric requires: 'User not found. Try search_users() with a partial name.' Pattern: recovery-guide.